Code with Jason

The episode contains real technical substance - spec-driven development for AI agents, property-based testing, TLA+ for distributed-systems verification, COE process mechanics, and the DynamoDB provisioned-throughput evolution - but this is badly diluted by a multi-minute snail-mail ad, extended Michigan/weather small talk, a power-tools brand tangent, and a lengthy oil/calculus analogy digression that consumes well over 15 minutes of a 70-minute runtime.

“we found that they would wander off. Maybe uh you'd come back and look at what it was doing, and it was doing some refactor that you didn't really want to do right now or ever”

“I think where people spend like the the the most productive time in in getting an agent to run um just better, is when they when they view everything that the agent did that they didn't want to do as a defect”

Spec-driven development for AI agents is increasingly common discourse, but the episode yields genuinely insider-original material: the internal history of Coral becoming Smithy/SDK generation, the AWS-wide weekly ops meeting where any team's dashboard can be pulled and interrogated live, and the feedback loop of mining COEs to improve the underlying service (e.g., adding DynamoDB auto-scaling after seeing throttling incidents in postmortems). These are non-recycled takes grounded in lived experience rather than thought-leadership clichés.

“There's no compression algorithm for experience.”

“we started DynamoDB with this notion of provisioned throughput because people wanted super predictable performance. So we would pre-provision capacity... but um, we didn't have auto-scaling at first. And so when we saw people run into uh, oh yeah, like it's it's hard to actually monitor this all the time and then dial it up. So let's just build that into the product”

David Yanacek is a genuine scale practitioner: he was on the pre-EC2 Amazon.com web server fleet, contributed to what became CloudWatch, worked on the Coral/Smithy service framework, DynamoDB, Lambda, and API Gateway, and is now building Kiro (agentic IDE) and the AWS DevOps Agent. This is rare hands-on depth across foundational cloud infrastructure. The score is not a perfect 20 because the conversation format prevents him from going as deep as his experience warrants.

“I've worked on Lambda then for managing servers, the building serverless, so you don't have to manage servers, API gateway because API routing is tedious and and easy to get wrong”

“we do a lot of formal verification methods, like TLA plus... we have things like DynamoDB, things like S3, where we have to prove that the replication, like these distributed systems around failover during a network partition while bootstrapping another replica”

The episode offers solid named specifics - TLA+ invariant verification for DynamoDB/S3 replication, Smithy as the open-source service definition language, P100 CPU percentile as the correct alarm metric, the five-whys Toyota method embedded in COEs, and the Coral framework becoming SDK generation - but lacks quantitative evidence: no reliability figures, no scale numbers, no timelines for product launches, and no data on agent productivity gains.

“I had an alarm, but I've alarmed on the median host CPU instead of really, it should always be the well, you should have that, but you should also have the the max, the P100, hundredth percentile host CPU to see like because maybe you're because it's really a distribution across your fleet”

“It's called Smithy, it's actually open source now, um, that defines services and what shapes and APIs and signing methods for identity”

The host has genuine technical knowledge and lands a few good moments - pushing back on postmortem efficiency and asking sharp questions about property-based testing - but frequently commandeers the conversation with lengthy personal analogies (oil/kerosene history, calculus derivatives, woodworking jigs, power-tool brands) that crowd out guest depth. There is no meaningful challenge to guest claims and the talk-time ratio heavily favors the host, limiting how far the practitioner's expertise is actually excavated.

“I don't know, David. This all sounds like it takes a lot of time. That doesn't sound very efficient.”

“I I I have to take us on a small digression. Um, my friend uh Stephen Baker mentioned the other day, uh he he made an analogy with oil. So uh back in the I don't know, mid-late 1800s, second half of the 1800s, um oil wasn't used for gasoline”